Abstract: The accuracy of spatial calibration in Stereo Particle Image Velocimetry (SPIV) has a great influence on the accuracy of the velocity measurement. In order to study the ability of various calibration model to deal with input error, a dimensionless parameter, namely, error attenuation coefficient, is defined to evaluate the response of spatial calibration model to input error. Based on this error attenua-tion coefficient, the error propagation characteristics of conventional spatial calibration model, including polynomial model and cam-era pinhole model, can be evaluated quantitatively. A neural network-based space calibration model is then developed. This new mod-el is naturally adaptive to multiple-camera joint calibration, which is lacked by conventional calibration models, thus is suitable for SPIV. Using synthetic experiment, it is demonstrated that this neural network model has the ability of suppressing the propagation of the input error in a large measurement parameter space, which are not posed by polynomial model or pinhole model. Additionally, it outbids traditional models in the scenario of high optical distortion. Therefore, this neural network model might be an ideal candidate for the spatial calibration of SPIV. Finally, it is confirmed in the experiment that the error of neural network calibration model is only a quarter of that of traditional model.

Key words: SPIV, Neural Network, Camera Calibration, Machine Vision, Spatial Positioning